Large transaction processing systems are typically distributed systems formed from a plurality of nodes interconnected by a network called an external switching fabric. One disadvantage of this type of distributed systems is that if a node becomes inoperable or inaccessible for a period of time, transactions that are processed by the node are disrupted. Inaccessibility may occur when software operating on the node undergoes a major overhaul or if there is a communication failure in the external switching fabric that isolates a node from the other nodes in the cluster. When such a failure occurs, it is desirable to make the node failure relatively unobservable to the users of the system. One existing solution, sometimes referred to as a system pair, is to provide a spare system that acts as a hot standby to the failed node. However, this is exceedingly expensive if the inoperable or inaccessible node is large.